非抽样小波变换耦合Zelnick矩的图像伪造检测算法  

Image forgery detection algorithm based on non-sampling wavelet transform coupled Zernike moment

作　　者：冯莉[1] 龚子华 文远保[3] 

机构地区：[1]厦门城市职业学院人文艺术系,福建厦门361008 [2]中船重工集团第七二二研究所,湖北武汉430000 [3]华中科技大学计算机科学与技术学院,湖北武汉430000

出　　处：《计算机工程与设计》2016年第6期1588-1592,共5页Computer Engineering and Design

摘　　要：为解决目前复制粘贴伪造检测算法鲁棒性较差、检测精度低等不足,提出非抽样小波变换(UWT)耦合泽尼克矩(Zernike)的复制移动图像伪造检测算法。利用非抽样小波变换将输入图像分解为近似系数LL和详细系数HH,得出块与块间的相似性和差异度;划分重叠块,计算图像子块间距;根据LL相似性和HH差异性排序分类;引入泽尼克矩,构建距离矩阵,进行图像特征匹配,对图像进行伪造检测。实验结果表明,该算法具有很好的平移不变性和幅值旋转不变性,抗后处理性能好,精测精度高,漏检率和误检率低。Aiming at the poor robustness and low detection accuracy of the copy paste forgery detection algorithm,the copy-moving image forgery detection algorithm based on non-sampling wavelet transform coupled Zernike moment was proposed.The input image was decomposed into approximation coefficients LL and detail coefficients HH,the similarity and difference between the blocks were obtained.LL the HH were divided into overlapping blocks,the image sub block spacing was calculated.According to the LL band similarity and HH band difference,sorting was proceeded.Zernike moment was introduced,and distance matrix was constructed,image feature matching was proceeded,the image of forgery was then detected.Experimental results show that：this algorithm has translation invariance and rotation invariance of the amplitude and good performance of anti-posttreatment with high measuring accuracy and low false detection and missed detection rate.

关 键 词：非抽样小波变换 泽尼克矩 复制粘贴 图像伪造检测 鲁棒性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]